======F.A.Q.======

=====Before You Start=====

====How do I get an account with the CHPC?====

Complete the online application on the [[http://users.chpc.ac.za|CHPC user database]].

=====Starting at the CHPC=====

====How do I log into the CHPC supercomputers?====

See the [[guide:connect| User Guide]].

=====Shell & Login Problems=====

====I cannot login because my account has been expired!====

You have been locked out because you have exceeded the home directory quota.  Submit a  [[https://users.chpc.ac.za/helpdesk|helpdesk request]] to ask for you account to be unlocked so that you can remove excess data.  You have until the following Monday morning to so this.

There is a 45 GB quota for data stored in your home directory.  Please note that this is separate to your Lustre space.  You will normally find a "lustre" directory in your home directory, but this is simply a symbolic link that points to your directory on a completely separate file system.  To find the amount of data in your home directory, use the "du" command.  You can get help with the "--help" option:
<code>
du --help
</code>

The ''-h'' option will provide the result in "human" readable units.  To see how much data you have in your home directory in total:
<code>
du -h -d0
</code> 

To find the amount of data in each directory, drill down one additional level:
<code>
du -h -d1
</code>

:!: **Please ensure that you understand that each user has access to TWO separate storage spaces:** :!:
  - ''/home/jblogs'' (if your userid is jblogs), also addressable as ''$HOME'' or ''~/''  .  This directory is located on an NFS file system.  Do **not** submit runs that write data to this space, as it is slow and small.  There is a 45 GB quota and if you exceed that quota your account will be expired.
  - ''/mnt/lustre/users/jblogs''.  There is normally also a symbolic link ("shortcut") to it from your home directory, typically ''/home/jblogs/lustre''.  The Lustre space is large, fast and is intended to be used when running jobs.  There is no size quota, but unused files will be deleted after 90 days.



====My .bashrc file is ignored when I login!====

The non-interactive startup file ''.bashrc'' is not executed for interactive shells, only //''.profile''// is. To fix this, add the following line to the end of your ''.profile'' file:
<code lang="bash">
source .bashrc
</code>

====I've compiled and installed package XYZ in my home directory but when I type XYZ I get "command not found"====

You need to add the path of the ''bin/'' sub-directory for your installation directory to your PATH environment.  Add something like the following to your **''.bashrc''** file (also see the previous question):
<code lang="bash">
export PATH=/export/home/your_username/XYZ/bin:$PATH
</code>
NB: replace ''your_username'' with your own user name as allocated for your CHPC account and replace ''XYZ'' with the full name of the sub-directory you installed the package XYZ in.

We recommend ''opt/'' so that. for example, if you installed Kactus 3.1.4 in ''~/opt/kactus-3.1.4/'' (in your home directory, where the ''~'' refers to the root of your home directory) you would add to ''.bashrc'' this line:
<code lang="bash">
export PATH=/export/home/your_username/opt/kactus-3.1.4/bin:$PATH
</code>

====I've compiled and installed library PQR in my home directory but when I type XYZ I get "libPQR library not found"====

You need to add the PQR library location to your LD_LIBRARY_PATH using the following code in your ''.bashrc''
<code lang="bash">
export LD_LIBRARY_PATH=/export/home/your_username/opt/PQR/lib:$LD_LIBRARY_PATH
</code>
Where you need to replace ''your_username'' with your own user name as allocated for your CHPC account and replace ''PQR'' with the full name of the sub-directory you installed the PQR library in.

==== My job is stuck in the queue and not running. WHY??? ====

Start by getting the numbers and status of your jobs, using the following command:
<code>
qstat -awu jblogs
</code>
Please use your own user-id, not jblogs!  

Now get more information about job number 123456.sched01 with the command:
<code>
qstat -f 123456.sched01 
</code>
Please use the number of the stuck job, not the made-up number used in this example ...  Now peruse the output from the the ''qstat -f'' command carefully.  Near the end will be a line similar to this one:
<code>     comment = Not Running: Insufficient amount of resource: nodetype </code> ,which is most likely to provide a clue.  In most cases, the comment will indeed be "Insufficient amount of resource ...", which may simply indicate that the system is very busy, and there are not enough free compute nodes currently available.  However, it may also indicate that you have asked for something that the cluster does not have, such as:
  - More than 24 CPUs per node for haswell_reg compute nodes, or more than 56 for haswell_fat nodes
  - More MPI processes than CPUs
  - Too much memory.  Please see  [[https://wiki.chpc.ac.za/quick:start#overview32_832_cores|this explanation]] on how to specify the required amount of memory. 

Also check that your research programme still has a valid allocation.  The scheduler should reject job submissions if you do not have a valid allocation, but if the allocation expired after job submission but before job launch, the job may be stuck in queued mode indefinitely.